At the beginning of 2016, I joined Opentrons, a Y-Combinator backed startup that builds robots for biologists.
While working at Opentrons I worked on a number of engineering projects. The first project I worked on was an optimization of the company's marketing website - I reduced HTML and CSS by 80% and load times by 80% (from 5+ seconds to ~1 second) and fixed a number of cross browser bugs that lowered our bounce rate, raised our conversion rate, and decreased support tickets.
The next project that I worked on was an overhaul of the main user-facing app. This app allows users to connect, calibrate, and control their robot for automating the pipetting of tiny amounts of liquid. The old app's repository can be found here, and the new repository that contains the redo of the app that I led can be found here.
The original app was written in plain HTML, CSS, and Javascript, and was buggy and difficult to maintain. It communicated over an ethernet cable to a Raspberry Pi that contained the backend code for sending singals to a Smoothieboard which physically controlled the robot. The refactor that I did initally removed the need for a Raspberry Pi by bundling a Python executeable that contained the backend server into the frontned. The removal of the Raspberry Pi reduced startup times of the robot from 3-5 minutes to 5-10 seconds, and the removal of the Ethernet cable in favor of a serial connection reduced customer support tickets by around 50% (the serial connection was far more reliable and cross OS compatible).
I also packaged the frontend into an Electron app, which eliminated cross browser compatibility issues, allowed us to auto-update the app, and gave us access to some neat OS specific actions, like the right-click context menu and the filebar.
After dealing with the major issues of customers being unable to connect to their robot, slow startup times for the robot, and browser cross compatability issues, I next focused on a more complete rewrite of the app. This rewrite introduced a completely overhauled UI, made the app much easier to maintain and reason about, reduced code bloat, and added a number of integration and unit tests.
I rewrote the app using Vue, a wonderful frontend framework. This allowed me to split the UI up into components and manage state with Vuex, a Redux-like state management library. I wrote unit and integration tests for the Vue components and Electron modules in Chai, Mocha, Karma, and Spectron.
While rewriting the main app, I worked with a team of other engineers to rewrite the backend framework that allows users to write custom "Protocols" for science experiments. It basically is a Python API that allows users to write sequences of commands that control the robot. Below is an example of what this means.
p200.pick_up_tip(p200_rack['A1'])
p200.aspirate(100, plate['A1']).dispense(50, plate['D1'])
p200.distribute(100, plate.wells('A2'), plate.rows('1', to='12'))
p200.drop_tip(p200_rack['A1'])
This was an incredible project to work on for a number of reasons. For one, it was a large and complex Python project that communicated over serial connection to a Smoothieboard, meaning I got to learn a lot of new concepts that I never would have otherwised been exposed to. Another fascinating aspect of this project was the API design itself - there was an interesting challenge in writing a Python API for non-programmer scientists that had to be self explanatory and concise, yet expressive and powerful enough to actually carry out real science experiments.
I was at Opentrons for about a year, and I both learned an insane amount of information and contributed a ton of interesting code. Also, we got featured in a research paper by the Mayo Clinic! I suppose that's another item off the bucket list.